tribhuvanesh / knockoffnets Goto Github PK

Hello, I have read your paper, and I notice that the 'adaptive' strategy is interesting. But I found that you didn't implement the ‘adaptive’ strategy in your code. If you can tell me how to achieve this, I would appreciate it.

When training with different budgets in adversary/train.py, you do not reset the model each run. So if you want to train with a budget of 100 and then 200, you actually train with 100 and then 300 (100+200). Because the model has already seen the first 100 at the start of the second budget; the budget adds up.

model = zoo.get_net(model_name, modelfamily, pretrained, num_classes=num_classes)
model = model.to(device)
for b in budgets:
    #train model

Solution: Reset or redefine the model each iteration.

Hi, thanks for sharing the code! my victim model is Caltech256-pretrained resnet34 as you provided and my adversary dataset is imagenet. I wonder how should I select the budget? Is 10k enough? Or I should choose 60k as CUB-200 demo? Or do you have any advice?

Thanks

The adaptive policy is the main contribution of paper, why NotImplemented?

if params['policy'] == 'random':
        adversary = RandomAdversary(blackbox, queryset, batch_size=batch_size)
    elif params['policy'] == 'adaptive':
        raise NotImplementedError()
    else:
        raise ValueError("Unrecognized policy")